DFNN is constructed by adding a recurrent layer between the normalized layer and output layer of ANFIS, and backpropogation learning algorithm (BP) is given to DFNN.

• 通过在ANFIS的归一化层和输出层之间加入递归层，构成了动态模糊神经网络(DFNN)，并推导了基于BP的反传学习算法。